Estimating the distribution of men who have sex with men (MSM) in the population based on Internet samples

Published: August 1, 2008

Estimating the distribution of men who have sex with men (MSM) in the population based on Internet samples

Background: Behavioural surveys in post-industrialized countries indicate that since the turn of the century the Internet has become an indispensable “place” for recruiting sexual partners, especially among MSM. Internet samples may thus offer new opportunities to collect epidemiologically relevant data on MSM populations, alternative or complementary to traditional venue-based sampling methods.

Methods: Based on two larger behavioural surveys among MSM in Germany in 2006 and 2007, we explored the use of data on the place of residence to estimate the regional size of MSM populations. In connection with routine surveillance data on HIV and syphilis, we calculated MSM specific incidence and prevalence rates on regional and city level.

Results: The regional proportion of MSM in the general population can be derived by adjusting the proportional regional distribution of survey participants for the regional distribution of the male general population. In larger cities or central city areas in Germany MSM may be concentrated 3- to 6-fold, while in rural areas the proportion of MSM in the general population may only be half as high as the national mean value. HIV and syphilis diagnosis incidence rates in MSM in 2006 in the largest cities ranged between 2 and 8/1,000 MSM for HIV and between 3 and 7/1,000 MSM for syphilis. Estimated HIV prevalence rates among MSM in the 16 federal states of Germany range between 1.5% and 9.2%.

Conclusion: Internet samples of MSM can be used as a low cost tool to estimate regional proportions of MSM in the general population. Together with routine surveillance data, incidence and prevalence rates of HIV and other reportable infections in MSM populations can be estimated with reasonable accuracy. Limitations of data on geographical allocations and potential biases of the sample composition have to be considered in the interpretation.

-Full abstract available at link below-

Leave a Reply